INDEX

Explanations

direct address names

The neuron flags personal names in email salutations (e.g. the addressee’s name following “Hi”).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

to

-1.71

in

-1.55

by

-1.41

on

-1.35

for

-1.32

 until

-1.32

as

-1.31

 without

-1.27

or

-1.23

at

-1.23

POSITIVE LOGITS

 réinitial

1.55

￬

1.37

 directement

1.32

 lavable

1.31

 reager

1.29

 VIET

1.28

 véhic

1.26

 lät

1.25

 PICTOGRAM

1.24

 desafios

1.23

Activations Density 0.010%